|Publication number||US5001766 A|
|Application number||US 07/194,686|
|Publication date||Mar 19, 1991|
|Filing date||May 16, 1988|
|Priority date||May 16, 1988|
|Also published as||CA1316461C|
|Publication number||07194686, 194686, US 5001766 A, US 5001766A, US-A-5001766, US5001766 A, US5001766A|
|Inventors||Henry S. Baird|
|Original Assignee||At&T Bell Laboratories|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (5), Referenced by (166), Classifications (25), Legal Events (4)|
|External Links: USPTO, USPTO Assignment, Espacenet|
The invention relates to the field of document reproduction, optical character readers, optical scanners, document readers and the like. In particular, the invention relates to the detection of and correction for rotational error (skew) between the dominant orientation of lines of printed objects, such as characters, of a document and a reference line observed by a reader, or the like, as being zero rotational error.
A fundamental and pervasive problem arises in the art of automatic document analysis, character/object recognition and related fields. The problem is the recognition and correction for skew in printed documents during the automated process. By skew, it is meant the angle between the dominant orientation of lines of characters or other textual objects of a document and a reference line observed by a reader, or the like, as representing zero angular error. Exemplary of the actions that are performed during document analysis is the segmentation of images of printed documents into blocks and lines of objects, such as characters. Known segmentation methods include those called top-down methods and bottom-up methods. Top-down methods characteristically operate by estimating some global properties of an image and by using the properties to guide segmentation into local regions whose local properties are estimated in turn. Bottom-up methods of segmentation characteristically operate by first clustering characters into lines, then lines into paragraphs, and so on. Unfortunately, the top-down methods tend to be excessively sensitive to non-zero skew.
A representative bottom-up method is described by Nagy, G. et. al. in an article entitled "Document Analysis with an Expert System," Proceedings, Pattern Recognition in Practice, Amsterdam, 1985. This bottom-up method relies on good skew alignment, with skew angle restricted to no more than a few degrees. However, while bottom-up methods are less sensitive to skew than top-down methods, the bottom-up methods are generally slower and suffer from other problems unrelated to skew sensitivity as well.
Hashizume, Yeh, & Rosenfeld, in an article entitled "A Method of Detecting the Orientation of Aligned Components", Pattern Recognition Letters, 1986, pp. 125-132, describe a skew determining method based on the premise that objects, e.g. characters, are often closer to one another along a dominant line orientation than in other directions. This technique computes the nearest neighbor of each object and connects each neighboring pair with a straight line segment. A PG,3 histogram of the orientations of these line segments is computed. The histogram may have a strongly-marked peak at the dominant skew angle. The skew angle is computed as an average of values near the peak. Among the known examples reported using this technique, the average error was 1.5 degrees and the worst 4.1 degrees.
W. Postl describe experiments with two methods of skew determining in a paper "Detection of Linear Oblique Structures and Skew Scan in Digitized Documents", Proceedings, Eighth International Conference on Pattern Recognition, Paris, October 1986, pp. 687-689. The first method applies the discrete two-dimensional Fourier transform to an image plane and examines a half plane of the power spectrum coefficients. The technique assumes an orientation angle and measures the energy in spatial frequencies at that orientation angle. The accuracy obtained with this method is not known. The second method similarly hunts for the maximum of a measure over a range of angles. The integral density of points is computed along assumed scan angles. For each pair of neighboring scan lines, the difference of their densities is computed. Finally, the sum of squares of these differences is computed.
Rastogi & Srihari describe a method using a Hough transform in an article, "Recognizing Textual Blocks Using the Hough Transform", Department Computer Science, University of Buffalo (SUNY), 1986. For each angle in a discrete representation of Hough space, the number of large "low-high-low transitions" is counted, and the maximum count is interpreted as identifying the dominant skew. In the five examples shown in the paper, skew angle was coarsely quantized in increments of 15 degrees.
While the above methods operate satisfactorily in some contexts, they are both slow and complex, and give coarse estimates of skew.
The invention is a method and apparatus for determining a predominant angle of orientation of an image with respect to a reference angle. A file of picture elements is generated which depicts the image with respect to the reference angle. The picture elements are projected onto a plurality of contiguous segments of imaginary lines at selected angles across the file. Each imaginary line is perpendicular to its associated direction of projection. The number of picture elements that fall into the segments for each projection are counted. An enhancement function is applied to the segment counts of each projection. The projection that generates the largest value of the enhancement function defines the angle of orientation of the image.
In a preferred embodiment, projections are first taken at a plurality of relatively coarse angles on both sides of the reference angle. This results in a first coarse estimate of the correct orientation. Then, projections are taken at more refined angles on both sides of the coarse first estimate. Also, the image file is compressed before further processing is performed, although this is not essential. The compressing involves performing a connected components analysis on the raw image file to locate individual image objects and by representing each object with one or more (preferably one) picture element. Processing operations are then performed on the compressed file.
The method works well for many page layout styles, including multiple columns, sparse tables, variable line spacings, mixed fonts, and a wide range of font styles and text sizes. Runtime of the method by a computer is proportional to the function n log (l/r), where n is the number of characters in the document and r is the angular resolution desired. A resolution of two minutes of arc is routinely achievable.
FIG. 1 shows a simplified block diagram of a document scanner, including a scanning platform, a rotatable scanning camera and controller for compensating for document skew;
FIG. 2 shows a slightly different embodiment of the system of FIG. 1, in which the platform is rotatable;
FIG. 3 conceptually illustrates the technique of the invention as aid to understanding; and
FIGS. 4 through 8 are flowcharts disclosing the principal steps of the invention.
FIG. 1 shows an illustrative document reader used to practice the invention. A document is placed automatically or manually on a transparent reading platform 10 where it is examined by a scanning device 12 in response to a READY signal arriving at a controller 14. The output of scanning device 12 is a stream of on/off pixels which reflect the document image on the platform 10 to the desired degree of resolution. Controller 14 processes the output from the scanning device in accordance with the algorithm described herein to determine the skew angle of the document. A signal is then applied to a motor 16 to rotate the scanning device 12 by an amount equal to the skew angle, thereby correcting for the document skew. Alternatively, motor 16 can be arranged to rotate the reading platform 10 rather than the scanning device 12 if desired, as shown in FIG. 2.
FIG. 3 pictorially illustrates the technique used to find the skew of a document. On the left side of FIG. 3 is shown a document skewed by some angle A with respect to an angle of reference observed by the scanner 12. The initial image file transmitted from the scanner 12 to the controller 14 reflects this skew. After correction for skew, the document has a skew angle substantially equal to zero within any desired tolerance down to about 2 minutes of arc.
The skew finding technique consists of first performing a connected components analysis on the image file to locate separate printed objects, such as characters, of the document. FIG. 3 shows two such objects "B" and "C" as part of the image file. Connected components are called "blobs" in various comments of the source code, which is discussed below and disclosed in Appendices 1 and 2. Each connected component is replaced with a single reference point in the file. This is illustrated in FIG. 3 by a dot at the lower rightmost point of the connected components B and C. The lower right corner of a connected component is preferably chosen for three reasons. First, in English text, character baselines are usually more uniform than top-lines due to the relatively low frequency of descenders. Second, if skew angle is not far from a zero reference angle, then any bottom point is a good approximation to the true base height of the character. Finally, the lower right corner is easy to compute. However, other reference points, such as the lower midpoint, may also be selected.
The purpose of a connected components analysis is to determine those groups of black pixels of the image file that touch one another directly or indirectly, such as an isolated character. The source code for performing a connected components analysis is not disclosed in the source code. This is because the code disclosed is tailored for a mainframe environment in which the connected components analysis is performed separately. Once the connected components analysis has been performed, the reference points replace the associated connected components in the image file for further processing. The connected components analysis is a conventional technique and is not described further herein. The reader is referred to "Connected Components in Binary Images: the Detection Problem", Ronse, C. & Devijver, P., Research Studies Press (Letchworth, Hertfordshire, England, 1984), for details of one technique of performing such an analysis.
The reference points are projected at a number of angles into a plurality of contiguous segments, which we call "bins", which partition an imaginary accumulating line. For each projection, the number of reference points that fall into each bin are counted. An example projection at 30 of FIG. 3 is at zero degrees with respect to the skewed document. The bin counts are shown to the right of 30. A second example projection is shown at 32. This projection is taken at the skew angle A. The associated bin counts are shown to the right of projection 32. As illustrated by FIG. 3, at angles close to the correct skew angle, reference points that are in alignment accumulate in a small number of bins, so that the distribution of counts concentrates more into extremely large and small values. Methods for computing and exploiting a numerical measure of this behavior of the distribution of counts are important features of this invention.
If the size of the bins (their length of their segments along the imaginary accumulating line) is too small, then, even at correct alignment, noise can scatter the projected points so that only a few of the bin counts are larger than one. If the bins are too large, then small angular changes have little effect. The document or the capabilities of the particular scanning system should be taken into account. Preferably, the bin size is chosen to be a fraction of the smallest point size expected to occur. For a system that handles text down to about 6 points, the bin size is preferably chosen to be 1/3 of the 6-point text (5 pixels at a digitizing resolution of 300 pixels/inch).
A superlinear enhancement function, described below, is applied to each set of bin counts. As illustrated in FIG. 3, the enhanced distribution forms a dominant peak which is at its largest value at the correct skew angle. It is this peak that is used to determine the correct skew.
The controller 14 may be implemented in any number of ways. For example, it might be a circuit composed of discrete components, a custom integrated circuit chip, a microprocessor chip driven by firmware or software or a more conventional type of computer. For this discussion, it is assumed that controller 14 is a microprocessor driven by firmware contained in a ROM. The program contained within the firmware is disclosed below with respect to the flowcharts in FIGS. 4 through 8 and source code listings in Appendixes 1 and 2. The source code is written in the C programming language. Appendix 1 contains the code for the header information required by the C language and the code for the main flow of the program. Appendix 2 contains the remaining code. This language is described in many text books, including The C Programming Language, Kernighan and Ritchie, Prentice-Hall, Inc. (1978). Only the major aspects of the source code will be explained in detail, as any programmer skilled in the C language is able to discern the remaining details from the code. The source code as disclosed in the Appendixes is arranged for execution on a mainframe computer under the control of an operator at a terminal. It will be discussed in this context for consistency. Modifications to the source code to adapt it to the environment of FIGS. 1 and 2 will be obvious to a skilled programmer.
Every C program is composed of a main function MAIN() and usually a number of other subfunctions or subroutines. A flowchart of MAIN() is shown in FIG. 4. The corresponding source code of MAIN() is disclosed at lines 262 through 290 of Appendix 1. At step 40 of FIG. 4, MAIN() first calls a subroutine PARSE-- ARGS(). PARSE-- ARGS() examines an input command describing a file of reference points to process for skew. The reference points may be the black picture elements of the entire, or may have been selected by a connected components analysis. The command may also contain various optional arguments to select parameters such as skew tolerance limits. PARSE-- ARGS() is disclosed at lines 143 through 260 of Appendix 1. Once the file of reference points to be processed is identified, MAIN() executes a subroutine READ-- POINTS() to input that file at step 42 of FIG. 4 into main memory for processing. READ-- POINTS() is disclosed at lines 292 through 326 of Appendix 1.
Step 44 calls a subroutine FIND-- SKEW(), which determines the skew of the set of reference points, as discussed below. FIND-- SKEW() returns a value indicating the skew angle to MAIN() in variable "skew" at line 286 of Appendix 1. Step 48 then compensates for the skew. This compensation takes the form of a mere printout at line 287 of MAIN(). In the system of FIGS. 1 and 2, the printout command "fprintf" at line 287 would be replaced with an output command to rotate the document platform 10 or the scanner 12.
The algorithm can be summarized as follows. It first probes at angles of coarse resolution until it detects the characteristic shape of the principal peak, and then refines the location of the peak's maximum by probes at finer resolution. Each probe computes the skew at a given angle, maintains a history of the probes, sorted by angle, and remembers the angle in variable SK.BEST.T whose skew value is largest. In the source code shown in Appendix 2, the user can suggest a preferred starting angle (variable SKEWO) (default is 0 degrees), an initial coarse resolution (variable RESO) (default is 0.5 degrees), and the finest resolution desired (variable RESF) (default is 0.0167 degrees). Probes are made at an initial angle SKEWO and at increments away from the initial angle at RESO, until a pattern emerges in which the present best angle SK.BEST.T is bracketed on both sides by at least three probes whose enhancement function values rise monotonically towards it. This completes the coarse location of the peak.
The location of the peak is then refined by an iterative procedure, each step of which is as follows. Below the present angle SK.BEST.T, a monotonically increasing subsequence of probes is selected whose enhancement function values rise monotonically. From this set of probes, no fewer than 3 and no more than 5 are selected to represent the "left slope" of the peak. Similarly, above the angle SK.BEST.T, a monotonically increasing subsequence of probes is selected whose enhancement function values fall monotonically. From this set of probes, no fewer than 3 and no more than 5 are selected to represent the "right slope" of the peak. Initially, those nearest to the present angle SK.BEST.T are chosen, but for later iterations (k), as probes cluster closely about the angle SK.BEST.T, the nearest k probes on each side of SK.BEST.T are ignored and farther ones chosen. These better describe the overall shape of the slopes near the peak. Smooth approximating functions are fitted to the left and the right slopes. The intersection of these approximating functions is computed and becomes the center of a new set of probes at a new, finer resolution C-- RESOL=C-- RESOL/2. The above described step is iterated until C-- RESOL falls below RESF. When this occurs, SK.BEST.T, the angle of the best probe so far, is returned as the PEAK angle.
FIND-- SKEW() is flowcharted in FIG. 5. The source code is shown at lines 528 through 679 of SKEWLIB.C in Appendix 2. Step 50 of FIG. 5 establishes a memory array into which angular projections across the document file can be made. This occurs beginning at line 559 of Appendix 2. Initially, a coarse resolution probing of the document file at five initial probing angles is performed. This is indicated as step 52 in FIG. 5 and begins at line 600 of Appendix 2. The actual probing is performed by a subroutine PROBE(), which is illustrated in FIG. 6. The source code for PROBE() is shown in Appendix 2 at lines 495 through 526. PROBE() is called five times for the initial angles at lines 606 through 610 of Appendix 2. The initial angles include one at SKEWO degrees with respect to the document file and four other at coarse angles spaced at RESO degree increments on either side of SKEWO. Step 60 of PROBE() executes a subroutine PROJECT(). PROJECT() is shown at lines 235 through 459 of Appendix 2. PROJECT() performs a projection at an angle contained in a variable "SKEW". The projection occurs at lines 235 through 432 of Appendix 2. The bin counts are accumulated in SK.PROJ at lines 332 through 361.
An enhancement function is applied to the bin counts at lines 416 through 418. Given a set of bin counts c(i), where i is an index ranging from one to the number of bins, the enhancement function computes the sum of E(c(i)) over all i, where the function E(c) is a "superlinear" function of c. That is, the value of E(c) grows asymptotically faster than that of any linear function of c. Many such functions exist which appear to work satisfactorily in practice, for example, E(c)=c log c, E(c)=c*c, and E(c)=c*c*c. Of these, E(c)=c*c is preferred, however, because it is well-motivated theoretically and it is relatively easy to compute.
After the initial five projections are obtained, a subroutine INFER-- PEAK() is iteratively called at line 614 of Appendix 2. This is shown at step 54 of FIG. 5. The iterations are symbolized by the flow line 56. INFER-- PEAK() is repetitively called, each time with a new incremental angle RESO degrees removed from the last call, until enough probes have been obtained to define two slopes of the peak.
After the coarse location of the peak is found, this location is refined by a series of iterations of probes taken at finer resolution. These iterations occur at lines 621 through 668 of Appendix 2. As part of these iterations, PROBE() and INFER-- PEAK() are iteratively executed with smaller incremental angles C-- RESOL to further refine the search around the skew angle SK.BEST.T. This is shown as step 58 and flow line 60 in FIG. 5. Eventually, FIND-- SKEW() returns with the best found skew angle in the variable SK.BEST.T. This occurs at line 678 of Appendix 2.
The flow chart of INFER-- PEAK() is shown in FIG. 7. The code is shown at lines 90 through 233 of Appendix 2. At steps 70 and 72 of FIG. 7, the representative probes for the left and right slopes are selected (Appendix 2, lines 132-147 and 148-169, respectively). Using this information, a least-squares fitting procedure is used at step 74 (Appendix 2, lines 170-217) to approximate the left and right slopes. The intersection point of the left and right slope approximating functions is calculated at step 76 (Appendix 2, lines 218-231). The angle SK.BEST.T is returned to the calling routine in variable PEAK.
The flowchart of PROBE() is shown in FIG. 6. The source code is shown in lines 495-526 of Appendix 2. PROBE() calls a function PROJECT() at step 60. The flowchart of PROJECT() is shown in FIG. 8. The source code is shown in lines 235-432 of Appendix 2. PROJECT() initializes a projection array at step 80. At step 82, the reference points of the image file are projected at an angle contained in variable SKEW. An enhancement function is applied to the bin counts resulting from the projection at step 84 and the value of the enhancement function is returned to PROBE().
The skew-finding algorithm has been applied to many documents representative of a wide variety of typographical and layout styles, having been selected from books, journals, theses, and typewritten pages. The selected documents include multiple columns, mixed fonts at various sizes, headers, trailers, and footnotes. They all possessed a dominant skew angle. The algorithm detected skew as small as 1/5 of the height of the text characters. The algorithm is able to compute skew to an accuracy as great as 2 minutes.
Most documents require about 40 probes for convergence to two minutes of arc. Runtime for documents with over 1000 characters is dominated by probing, which in turn is dominated by projecting. The iterative probing technique is thus about a factor of 100 times faster than an exhaustive search over all angles at the desired resolution.
It is to be understood that the above described arrangements are merely illustrative of the application of principles of the invention and that other arrangements may be devised by workers skilled in the art without departing from the spirit and scope of the invention. ##SPC1##
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US4338588 *||Jun 2, 1980||Jul 6, 1982||International Business Machines Corporation||Arrangement for determining the optimum scan angle for printed documents|
|US4497066 *||Mar 5, 1982||Jan 29, 1985||Texas Instruments Incorporated||Video data acquisition system|
|US4539703 *||Mar 5, 1982||Sep 3, 1985||Texas Instruments Incorporated||Video data acquisition system and hand-held application module for operation thereof|
|US4558461 *||Jun 17, 1983||Dec 10, 1985||Litton Systems, Inc.||Text line bounding system|
|US4688088 *||Apr 15, 1985||Aug 18, 1987||Canon Kabushiki Kaisha||Position detecting device and method|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US5149977 *||Feb 21, 1991||Sep 22, 1992||Eastman Kodak Company||Document reader apparatus|
|US5187753 *||Dec 8, 1989||Feb 16, 1993||Xerox Corporation||Method and apparatus for identification and correction of document skew|
|US5235651 *||Aug 6, 1991||Aug 10, 1993||Caere Corporation||Rotation of images for optical character recognition|
|US5245676 *||Dec 21, 1989||Sep 14, 1993||Xerox Corporation||Determination of image skew angle from data including data in compressed form|
|US5285504 *||Sep 27, 1991||Feb 8, 1994||Research Foundation Of The State University Of New York||Page segmentation with tilt compensation|
|US5287199 *||Feb 27, 1992||Feb 15, 1994||At&T Bell Laboratories||Facsimile message processing and routing system|
|US5355420 *||Oct 19, 1992||Oct 11, 1994||Xerox Corporation||Method and apparatus for identification of document skew|
|US5410417 *||Feb 10, 1994||Apr 25, 1995||Motorola, Inc.||Method and apparatus for correcting an angle of an optical image for improving the efficiency of facsimile encoding of the image|
|US5414781 *||Nov 24, 1993||May 9, 1995||Xerox Corporation||Method and apparatus for classifying documents|
|US5416851 *||Jul 30, 1991||May 16, 1995||Xerox Corporation||Image analysis based on location sampling|
|US5487115 *||Aug 17, 1993||Jan 23, 1996||United Parcel Service||Method and apparatus for determining the fine angular orientation of bar code symbols in two-dimensional CCD images|
|US5506918 *||Dec 23, 1992||Apr 9, 1996||Kabushiki Kaisha Toshiba||Document skew detection/control system for printed document images containing a mixture of pure text lines and non-text portions|
|US5513277 *||Jun 14, 1993||Apr 30, 1996||Xerox Corporation||Measuring character and stroke sizes and spacings for an image|
|US5513304 *||Feb 4, 1994||Apr 30, 1996||Xerox Corporation||Method and apparatus for enhanced automatic determination of text line dependent parameters|
|US5517587 *||Sep 23, 1994||May 14, 1996||International Business Machines Corporation||Positioning method and apparatus for line scanned images|
|US5579414 *||May 22, 1995||Nov 26, 1996||Fast; Bruce B.||OCR image preprocessing method for image enhancement of scanned documents by reversing invert text|
|US5583956 *||Jun 14, 1994||Dec 10, 1996||The Board Of Trustees Of The Leland Stanford Junior University||Estimation of skew angle in text image|
|US5588072 *||Dec 22, 1993||Dec 24, 1996||Canon Kabushiki Kaisha||Method and apparatus for selecting blocks of image data from image data having both horizontally- and vertically-oriented blocks|
|US5590224 *||May 22, 1995||Dec 31, 1996||Fast; Bruce B.||OCR image preprocessing method for image enhancement of scanned documents by correction of registration|
|US5594814 *||May 22, 1995||Jan 14, 1997||Fast; Bruce B.||OCR image preprocessing method for image enhancement of scanned documents|
|US5594815 *||May 22, 1995||Jan 14, 1997||Fast; Bruce B.||OCR image preprocessing method for image enhancement of scanned documents|
|US5594817 *||May 22, 1995||Jan 14, 1997||Fast; Bruce B.||OCR image pre-processor for detecting and reducing skew of the image of textual matter of a scanned document|
|US5625719 *||May 22, 1995||Apr 29, 1997||Fast; Bruce B.||OCR image preprocessing method for image enhancement of scanned documents|
|US5627912 *||Dec 6, 1995||May 6, 1997||Yozan Inc.||Inspection method of inclination of an IC|
|US5692069 *||Mar 17, 1995||Nov 25, 1997||Eastman Kodak Company||Apparatus for performing character segmentation using slant histograms|
|US5696841 *||Oct 12, 1993||Dec 9, 1997||Canon Kabushiki Kaisha||Image processing method and apparatus for extracting vertically and horizontally written text|
|US5729635 *||Oct 17, 1996||Mar 17, 1998||Tmssequoia||OCR image free-processing method for image enhancement of scanned documents|
|US5737444 *||Jan 13, 1997||Apr 7, 1998||Ministero Dell `Universita` e Della Ricerca Scientifica e Technologica||Binary digital image feature extracting process|
|US5825920 *||Dec 22, 1992||Oct 20, 1998||Hitachi, Ltd.||Method and unit for binary processing in image processing unit and method and unit for recognizing characters|
|US5832110 *||May 28, 1996||Nov 3, 1998||Ricoh Company, Ltd.||Image registration using projection histogram matching|
|US5854853 *||Feb 5, 1996||Dec 29, 1998||Canon Kabushika Kaisha||Method and apparatus for selecting blocks of image data from image data having both horizontally- and vertically-oriented blocks|
|US6009198 *||Nov 21, 1997||Dec 28, 1999||Xerox Corporation||Method for matching perceptual shape similarity layouts across multiple 2D objects|
|US6282324 *||Aug 31, 1995||Aug 28, 2001||Northrop Grumman Corporation||Text image deblurring by high-probability word selection|
|US6285802||Apr 8, 1999||Sep 4, 2001||Litton Systems, Inc.||Rotational correction and duplicate image identification by fourier transform correlation|
|US6310984 *||Apr 9, 1998||Oct 30, 2001||Hewlett-Packard Company||Image processing system with image cropping and skew correction|
|US6373997||Mar 29, 1993||Apr 16, 2002||Xerox Corporation||Coarse and fine skew measurement|
|US6493470 *||Jun 19, 1996||Dec 10, 2002||Canon Kabushiki Kaisha||Image processing method and apparatus for detecting the tilt amount of input image data|
|US6549680 *||Jun 23, 1998||Apr 15, 2003||Xerox Corporation||Method and apparatus for deskewing and despeckling of images|
|US6683983||Feb 29, 2000||Jan 27, 2004||Riso Kagaku Corporation||Document-inclination detector|
|US6768820||Sep 11, 2000||Jul 27, 2004||Agilent Technologies, Inc.||Method and system for extracting data from surface array deposited features|
|US6771842 *||Sep 19, 2000||Aug 3, 2004||Fujitsu Limited||Document image skew detection method|
|US6941030 *||Nov 30, 2000||Sep 6, 2005||Pfu Limited||Character-recognition pre-processing apparatus and method, and program recording medium|
|US6985640||Jan 7, 2002||Jan 10, 2006||Xerox Corporation||Parallel non-iterative method of determining and correcting image skew|
|US7133573 *||Nov 26, 2002||Nov 7, 2006||Eastman Kodak Company||Document skew accommodation|
|US7142727||Dec 17, 2002||Nov 28, 2006||Xerox Corporation||Non-iterative method of calculating image skew|
|US7145699||Oct 18, 2002||Dec 5, 2006||Sharp Laboratories Of America, Inc.||System and method for digital document alignment|
|US7151860 *||Jul 7, 2000||Dec 19, 2006||Fujitsu Limited||Document image correcting device and a correcting method|
|US7221774 *||Mar 28, 2003||May 22, 2007||Canon Kabushiki Kaisha||Local phase filter to assist correlation|
|US7305612 *||Mar 31, 2003||Dec 4, 2007||Siemens Corporate Research, Inc.||Systems and methods for automatic form segmentation for raster-based passive electronic documents|
|US7330606||Jul 19, 2004||Feb 12, 2008||Agilent Technologies, Inc.||Method and system for extracting data from surface array deposited features|
|US7391930 *||Dec 17, 2004||Jun 24, 2008||Primax Electronics Ltd.||Angle de-skew device and method thereof|
|US7627191||Aug 14, 2006||Dec 1, 2009||Xerox Corporation||Image manipulation|
|US7877213||Sep 22, 2005||Jan 25, 2011||Agilent Technologies, Inc.||System and methods for automated processing of multiple chemical arrays|
|US8055098||Jan 26, 2007||Nov 8, 2011||Affymetrix, Inc.||System, method, and product for imaging probe arrays with small feature sizes|
|US8160393 *||Sep 15, 2009||Apr 17, 2012||Certifi Media Inc.||Method for image skew detection|
|US8233735||Sep 17, 2008||Jul 31, 2012||Affymetrix, Inc.||Methods and apparatus for detection of fluorescently labeled materials|
|US8380573||Jul 22, 2008||Feb 19, 2013||Cummins-Allison Corp.||Document processing system|
|US8391583||Jul 14, 2010||Mar 5, 2013||Cummins-Allison Corp.||Apparatus and system for imaging currency bills and financial documents and method for using the same|
|US8396278||Jun 23, 2011||Mar 12, 2013||Cummins-Allison Corp.||Document processing system using full image scanning|
|US8417017||Apr 13, 2010||Apr 9, 2013||Cummins-Allison Corp.||Apparatus and system for imaging currency bills and financial documents and method for using the same|
|US8428332||Apr 13, 2010||Apr 23, 2013||Cummins-Allison Corp.||Apparatus and system for imaging currency bills and financial documents and method for using the same|
|US8433123||Apr 13, 2010||Apr 30, 2013||Cummins-Allison Corp.|
|US8433126||Sep 22, 2011||Apr 30, 2013||Cummins-Allison Corp.||Check and U.S. bank note processing device and method|
|US8437528||Apr 13, 2010||May 7, 2013||Cummins-Allison Corp.|
|US8437529||Apr 13, 2010||May 7, 2013||Cummins-Allison Corp.|
|US8437530||Apr 13, 2010||May 7, 2013||Cummins-Allison Corp.|
|US8437531||Sep 22, 2011||May 7, 2013||Cummins-Allison Corp.||Check and U.S. bank note processing device and method|
|US8437532||Apr 13, 2010||May 7, 2013||Cummins-Allison Corp.|
|US8442296||Sep 22, 2011||May 14, 2013||Cummins-Allison Corp.||Check and U.S. bank note processing device and method|
|US8459436||Dec 10, 2012||Jun 11, 2013||Cummins-Allison Corp.||System and method for processing currency bills and tickets|
|US8467591||Apr 13, 2010||Jun 18, 2013||Cummins-Allison Corp.|
|US8478019||Apr 13, 2010||Jul 2, 2013||Cummins-Allison Corp.|
|US8478020||Apr 13, 2010||Jul 2, 2013||Cummins-Allison Corp.|
|US8514379||Dec 11, 2009||Aug 20, 2013||Cummins-Allison Corp.||Automated document processing system and method|
|US8520976||Sep 23, 2011||Aug 27, 2013||Affymetrix, Inc.||System, method, and product for imaging probe arrays with small feature size|
|US8538123||Apr 13, 2010||Sep 17, 2013||Cummins-Allison Corp.|
|US8542904||Mar 7, 2013||Sep 24, 2013||Cummins-Allison Corp.|
|US8559695||Mar 5, 2013||Oct 15, 2013||Cummins-Allison Corp.|
|US8577174 *||Jun 24, 2010||Nov 5, 2013||Samsung Electronics Co., Ltd.||Image processing apparatus and method|
|US8594414||Mar 5, 2013||Nov 26, 2013||Cummins-Allison Corp.|
|US8620080||Sep 26, 2008||Dec 31, 2013||Sharp Laboratories Of America, Inc.||Methods and systems for locating text in a digital image|
|US8625875||Feb 22, 2012||Jan 7, 2014||Cummins-Allison Corp.||Document imaging and processing system for performing blind balancing and display conditions|
|US8627939||Dec 10, 2010||Jan 14, 2014||Cummins-Allison Corp.|
|US8639015||Mar 5, 2013||Jan 28, 2014||Cummins-Allison Corp.|
|US8644583||Feb 4, 2013||Feb 4, 2014||Cummins-Allison Corp.|
|US8644584||Mar 5, 2013||Feb 4, 2014||Cummins-Allison Corp.|
|US8644585||Mar 5, 2013||Feb 4, 2014||Cummins-Allison Corp.|
|US8655045||Feb 6, 2013||Feb 18, 2014||Cummins-Allison Corp.||System and method for processing a deposit transaction|
|US8655046||Mar 6, 2013||Feb 18, 2014||Cummins-Allison Corp.|
|US8701857 *||Oct 29, 2008||Apr 22, 2014||Cummins-Allison Corp.||System and method for processing currency bills and tickets|
|US8714336||Apr 2, 2012||May 6, 2014||Cummins-Allison Corp.|
|US8787652||Oct 21, 2013||Jul 22, 2014||Cummins-Allison Corp.|
|US8929640||Apr 15, 2011||Jan 6, 2015||Cummins-Allison Corp.|
|US8944234||Mar 11, 2013||Feb 3, 2015||Cummins-Allison Corp.|
|US8948490||Jun 9, 2014||Feb 3, 2015||Cummins-Allison Corp.|
|US8958626||Mar 11, 2013||Feb 17, 2015||Cummins-Allison Corp.|
|US9129271||Feb 28, 2014||Sep 8, 2015||Cummins-Allison Corp.||System and method for processing casino tickets|
|US9141876||Feb 22, 2013||Sep 22, 2015||Cummins-Allison Corp.||Apparatus and system for processing currency bills and financial documents and method for using the same|
|US9142075||Dec 23, 2013||Sep 22, 2015||Cummins-Allison Corp.|
|US9160885||Jul 2, 2009||Oct 13, 2015||Hewlett-Packard Development Company, L.P.||Skew detection|
|US9189780||Dec 24, 2014||Nov 17, 2015||Cummins-Allison Corp.||Apparatus and system for imaging currency bills and financial documents and methods for using the same|
|US9195889||Feb 4, 2015||Nov 24, 2015||Cummins-Allison Corp.||System and method for processing banknote and check deposits|
|US9355295||Mar 11, 2013||May 31, 2016||Cummins-Allison Corp.|
|US9390574||Jan 27, 2011||Jul 12, 2016||Cummins-Allison Corp.||Document processing system|
|US9445025||Aug 23, 2013||Sep 13, 2016||Affymetrix, Inc.||System, method, and product for imaging probe arrays with small feature sizes|
|US9477896||Jan 9, 2014||Oct 25, 2016||Cummins-Allison Corp.|
|US9495808||Jul 22, 2015||Nov 15, 2016||Cummins-Allison Corp.||System and method for processing casino tickets|
|US9508168||Dec 11, 2009||Nov 29, 2016||Xerox Corporation||Method and system for utilizing transformation matrices to process rasterized image data|
|US9509884 *||Sep 8, 2015||Nov 29, 2016||Hewlett-Packard Development Company, L.P.||Skew detection|
|US9558418||Aug 14, 2015||Jan 31, 2017||Cummins-Allison Corp.||Apparatus and system for processing currency bills and financial documents and method for using the same|
|US9626602||Dec 11, 2009||Apr 18, 2017||Xerox Corporation||Method and system for utilizing transformation matrices to process rasterized image data|
|US9626603||Dec 11, 2009||Apr 18, 2017||Xerox Corporation||Method and system for utilizing transformation matrices to process rasterized image data|
|US20020126904 *||Nov 30, 2000||Sep 12, 2002||Hiroshi Kakutani||Character-recognition pre-processing apparatus and method, and program recording medium|
|US20020173277 *||May 14, 2002||Nov 21, 2002||Ntt Docomo, Inc.||Mobile communication systems|
|US20030128895 *||Jan 7, 2002||Jul 10, 2003||Xerox Corporation||Parallel non-iterative method of determining and correcting image skew|
|US20030231768 *||Mar 28, 2003||Dec 18, 2003||Canon Kabushiki Kaisha||Local phase filter to assist correlation|
|US20040076341 *||Oct 18, 2002||Apr 22, 2004||Sharp Laboratories Of America, Inc.||System and method for digital document alignment|
|US20040101211 *||Nov 26, 2002||May 27, 2004||Eastman Kodak Company||Document skew accommodation|
|US20040114831 *||Dec 17, 2002||Jun 17, 2004||Xerox Corporation||Non-iterative method of calculating image skew|
|US20040194035 *||Mar 31, 2003||Sep 30, 2004||Amit Chakraborty||Systems and methods for automatic form segmentation for raster-based passive electronic documents|
|US20040196456 *||Apr 21, 2004||Oct 7, 2004||Affymetrix, Inc.||Method and apparatus for detection of fluorescently labeled materials|
|US20040243341 *||May 30, 2003||Dec 2, 2004||Le Cocq Christian A.||Feature extraction methods and systems|
|US20040264807 *||Jul 19, 2004||Dec 30, 2004||Yakhini Zohar H.||Method and system for extracting data from surface array deposited features|
|US20050281462 *||Jun 16, 2004||Dec 22, 2005||Jayati Ghosh||System and method of automated processing of multiple microarray images|
|US20060009917 *||Sep 6, 2005||Jan 12, 2006||Le Cocq Christian A||Feature extraction methods and systems|
|US20060111850 *||Dec 30, 2005||May 25, 2006||Affymetrix, Inc.||Method and apparatus for detection of fluorescently labeled materials|
|US20060133692 *||Dec 17, 2004||Jun 22, 2006||Yu-Ting Shih||Angle de-skew device and method thereof|
|US20060258002 *||Mar 3, 2006||Nov 16, 2006||Affymetrix, Inc.||Scanned image alignment systems and methods|
|US20070174007 *||Sep 22, 2005||Jul 26, 2007||Jayati Ghosh||System and methods for automated processing of multiple chemical arrays|
|US20070197911 *||Jan 26, 2007||Aug 23, 2007||Affymetrix, Inc.||System, Method, and Product for Imaging Probe Arrays with Small Feature Sizes|
|US20080037899 *||Aug 14, 2006||Feb 14, 2008||Xerox Corporation||Image manipulation|
|US20080285838 *||Jul 22, 2008||Nov 20, 2008||Cummins-Allison Corp||Document Processing System|
|US20090010524 *||Sep 17, 2008||Jan 8, 2009||Affymetrix, Inc.||Methods and apparatus for detection of fluorescently labeled materials|
|US20090087076 *||Oct 29, 2008||Apr 2, 2009||Cummins-Allison Corp.||System and method for processing currency bills and tickets|
|US20100067826 *||Sep 15, 2009||Mar 18, 2010||Certifi Media Inc.||Method for Image Skew Detection|
|US20100080461 *||Sep 26, 2008||Apr 1, 2010||Ahmet Mufit Ferman||Methods and Systems for Locating Text in a Digital Image|
|US20100092065 *||Dec 11, 2009||Apr 15, 2010||Cummins-Allison Corp.||Automated document processing system and method|
|US20100157319 *||Dec 11, 2009||Jun 24, 2010||Xerox Corporation||Method and system for utilizing transformation matrices to process rasterized image data|
|US20100157320 *||Dec 11, 2009||Jun 24, 2010||Xerox Corporation||Method and system for utilizing transformation matrices to process rasterized image data|
|US20100157321 *||Dec 11, 2009||Jun 24, 2010||Xerox Corporation||Method and system for utilizing transformation matrices to process rasterized image data|
|US20110142352 *||Jun 24, 2010||Jun 16, 2011||Samsung Electronics Co., Ltd.||Image processing apparatus and method|
|US20150042661 *||Aug 5, 2014||Feb 12, 2015||Kabushiki Kaisha Toshiba||Handwritten document processing device, handwritten document processing method, and handwritten document processing program|
|US20160050338 *||Sep 8, 2015||Feb 18, 2016||Hewlett-Packard Development Company, L.P.||Skew detection|
|CN100442215C||Sep 29, 2003||Dec 10, 2008||微软公司||Decoding and error correction in 2-d arrays|
|CN104346320A *||Aug 8, 2014||Feb 11, 2015||株式会社东芝||Handwritten document processing device, handwritten document processing method, and handwritten document processing program|
|EP0526195A2 *||Jul 30, 1992||Feb 3, 1993||Xerox Corporation||Image processors|
|EP0526195A3 *||Jul 30, 1992||Mar 16, 1994||Xerox Corp||Title not available|
|EP0526197A2 *||Jul 30, 1992||Feb 3, 1993||Xerox Corporation||Image processing|
|EP0526197A3 *||Jul 30, 1992||Apr 13, 1994||Xerox Corp||Title not available|
|EP0526199A2 *||Jul 30, 1992||Feb 3, 1993||Xerox Corporation||Image processing apparatus|
|EP0526199A3 *||Jul 30, 1992||Mar 16, 1994||Xerox Corp||Title not available|
|EP0621554A2 *||Apr 18, 1994||Oct 26, 1994||Xerox Corporation||Method and apparatus for automatic determination of text line, word and character cell spatial features|
|EP0621554A3 *||Apr 18, 1994||May 24, 1995||Xerox Corp||Method and apparatus for automatic determination of text line, word and character cell spatial features.|
|EP0684577A3 *||May 17, 1995||Sep 4, 1996||Mini Ricerca Scient Tecnolog||Binary digital image feature extracting process.|
|EP0703696A2||Sep 21, 1995||Mar 27, 1996||International Business Machines Corporation||Skew correction in line scanned images|
|EP1033864A2 *||Feb 28, 2000||Sep 6, 2000||Riso Kagaku Corporation||Document-inclination detector|
|EP1033864A3 *||Feb 28, 2000||Jul 25, 2001||Riso Kagaku Corporation||Document-inclination detector|
|EP1162572A2 *||Jun 6, 2001||Dec 12, 2001||Agilent Technologies, Inc. (a Delaware corporation)||Method and system for evaluating orientation of molecular array|
|EP1162572A3 *||Jun 6, 2001||Feb 5, 2003||Agilent Technologies, Inc. (a Delaware corporation)||Method and system for evaluating orientation of molecular array|
|EP1427181A2 *||Nov 17, 2003||Jun 9, 2004||Eastman Kodak Company||Document skew accommodation|
|EP1427181A3 *||Nov 17, 2003||Oct 20, 2004||Eastman Kodak Company||Document skew accommodation|
|EP1481727A2 *||May 27, 2004||Dec 1, 2004||Agilent Technologies Inc||Feature extraction methods and systems|
|EP1481727A3 *||May 27, 2004||Jan 12, 2005||Agilent Technologies Inc||Feature extraction methods and systems|
|EP1494171A1 *||Jun 6, 2001||Jan 5, 2005||Agilent Technologies, Inc.||Method and system for evaluating orientation of molecular array|
|WO1995022225A1 *||Feb 3, 1995||Aug 17, 1995||Motorola Inc.||Correcting angle of optical image for improving efficiency of facsimile encoding|
|WO2017020917A1 *||Jul 31, 2015||Feb 9, 2017||Hewlett-Packard Indigo B.V.||Detection of streaks in images|
|U.S. Classification||382/290, 382/177, 382/174|
|International Classification||G06K9/32, G06K9/46, H04N1/00, H04N1/04|
|Cooperative Classification||G06K9/3283, H04N1/00787, H04N1/00681, H04N1/00795, G06K2209/01, H04N2201/0432, H04N1/04, H04N1/00769, H04N1/00721, G06K9/4647|
|European Classification||H04N1/00G5C, H04N1/00G6B, H04N1/00G2F, H04N1/00G, G06K9/32S1, G06K9/46B1, H04N1/04, H04N1/00H|
|May 16, 1988||AS||Assignment|
Owner name: BELL TELEPHONE LABORATORIES INCORPORATED, 600 MOUN
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST.;ASSIGNOR:BAIRD, HENRY S.;REEL/FRAME:004892/0141
Effective date: 19880516
Owner name: BELL TELEPHONE LABORATORIES INCORPORATED, NEW JERS
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:BAIRD, HENRY S.;REEL/FRAME:004892/0141
Effective date: 19880516
|Jul 27, 1994||FPAY||Fee payment|
Year of fee payment: 4
|Aug 31, 1998||FPAY||Fee payment|
Year of fee payment: 8
|Aug 29, 2002||FPAY||Fee payment|
Year of fee payment: 12